I’ve been involved with Event Management and more recently AIOps products for over 30 years and in this time, I’ve written several presentations and blogs on how to secure a Return on Investment (ROI) for the software we are implementing or selling. Many of these cases rely on an understanding of the cost of an outage and a good understanding of the costs of managing day to day incidents which in several instances are not fully known or are based on guesswork. This is not impossible, but it can be prone to a large margin of error. The business case for Turbonomic is different. This is the first product I’ve looked at that is not only designed to improve the performance of your Virtual or Cloud estate but to save money while it is doing it. Turbonomic regularly achieves this which is evidenced by 88% of Turbonomic customers receiving a full ROI in 3 months or less through ongoing performance and efficiency gains, as well as the coincident capital (CapEx) and operational (OpEx) deferments that the solution enables.
In addition, Forrester Consulting conducted a Total Economic Impact™ (TEI) study and projected that Turbonomic delivered a 471% return on investment for a composite organization while paying for itself in less than six months. To do this study Forrester interviewed 5 customers across varying industries, geographies, and company sizes on the benefits, costs, and risks associated with their investment in Turbonomic. From this customer data, Forrester created a single composite organization to represent the projected outcomes of the first three years of using Turbonomic the results of which are shown below.
|Financial Summary (risk-adjusted estimates)|
|ROI||Payback Period (Months)||Total benefits (PV)||Total costs (PV)||Net present value|
You can look at the full report by clicking here.
The Turbonomic AI engine applies economic principles to create actions that manage application resourcing, so applications get the resources they need when they need them. In this blog, I’m going to concentrate on these procedures and describe specifically the actions that can save money and hopefully demonstrate how the ROI can be achieved in such a short time.
Why is it Needed?
Let’s start by looking at why a tool like Turbonomic is needed. Many companies over-spec their infrastructure because quite rightly their main aim is to reliably deliver the performance needed to make sure their customers’ needs are met. Turbonomic removes this requirement. The Turbonomic solution assures application response time continuously (especially during peak demands) but it enables users to achieve both performance and efficiency by identifying which workloads can and should be placed together, as well as how they should be sized, assure they have access to the resources they need to perform while maximizing VM densities. After it is installed 87% of Turbonomic customers can consolidate their virtual workloads by intelligently placing workloads while mitigating performance risk. The analysis considers all the conditions in the current environment and identiﬁes the optimal workload distribution for each cluster. This can include moving your current VMs to other hosts within the given cluster if this would result in a more desirable workload distribution.
How does it achieve the savings?
Turbonomic is an autonomic platform that collects real-time demand, topology, and telemetry data via REST APIs from across the entities in your virtualized and cloud infrastructure (AWS, Azure, Google Cloud). This includes hypervisors, storage controllers, network components, converged fabrics, load balancers, cloud regions and cloud platforms. It then uses this data to map the end-to-end relationships and interdependencies between them and uses this data to suggest actions that will keep your environment at its desired state. As demand changes (either up or down) Turbonomic will look at the best value instances and suggest changes whilst also supplying the cost of each action. The main actions associated with cost are:
- Scaling — Resize allocation of resources, based on profitability
- Resize up, shown as a Performance Action or Investment
- Resize down, shown as a Savings Action
- Start/Buy — Start a new Reserved Instance (RI) to add capacity to the environment, shown as a required investment.
- RI Optimization — Purchase RIs for specific workloads or move to RI tiers that are more appropriate for your applications’ requirements
- Delete — Remove storage (for example, datastores on disk arrays or unattached volumes)
Note: If you are worried that Turbonmoic will run the actions that may cause disruption, there is a mode (Enforce Non Disruptive Mode) that will post the action as Recommend if the action will potentially disrupt service (e.g. reboot the VM).
As I said earlier Turbonomic is primarily a performance tool and so any suggestions it provides uses this as the primary goal, however when these goals are met and there are also potential cost savings these are also suggested. The Scaling actions recommend capacity updates for items like Containers, VMs, Storage, etc. There are 2 types of scaling:
- Vertical Scaling – increase or decrease the capacity of resources on existing entities.
- Horizontal Scaling – provisioning of new providers.
Performance degradation or cloud cost overruns are more likely to occur when scaling to peaks or averages. Turbonomic uses percentile-based scaling to help customers achieve true cloud elasticity using configurable observation periods to ensure that Turbonomic is analyzing data that accounts for unique business cycles.
For the resulting actions, the list shows the current cost for the source workload, and the projected cost given the change (as shown in the screenshot below). The current cost is worked out using the actual costs for that workload, but the projected cost is an estimate based on average utilization for the VM, for the costs of the given tier. On the cloud, scaling actions change the VM to a different instance type such as changing a VM to an instance type with a different capacity or changing on-demand to a Reserved Instance (RI).
For storage, Turbonomic uses the same powerful analytics that determines the right compute scaling actions. Turbonomic considers IOPS and throughput, to determine when you need to do the following actions:
- Scale between cloud tiers (AWS & Azure) for performance (IOPS, throughput) and cost
- Size up volumes for performance (IOPS, throughput)
- Modify capacity of IOPS & throughput limit for Azure Ultra or IOPS limits for EBS io1 & io2.
With automated RI-aware scaling and purchase recommendations Turbonomic helps in the following ways:
- Helps you purchase the first order of reservations
- Optimizes the use of your current reservations
- Helps you re-purchase soon-to-be-expired reservations
- Continues reservation purchases for growing environments
Turbonomic can recommend that you start a suspended entity to add capacity to the environment or purchase Reserved Instance (RI) capacity to reduce costs for the current workload. For purchases, Turbonomic considers the count of workloads in a family, plus their hours of active-state condition, plus RI costs to arrive at the RI capacity that is suggested to purchase. To recommend placing workloads on Reserved Instances (RIs), Turbonomic uses the real pricing plans that are available to the targets public cloud accounts. You can add more detail to this by setting up an RI Purchase Profile.
As Turbonomic calculates actions to purchase RI capacity, it assumes that any other pending actions for the workload will also be executed. For example, assume a workload running on an r4.xlarge template. If Turbonomic recommends changing that instance type to an m5.medium, it can recommend that you purchase an m5 RI to cover the workload and reduce costs. This purchase could be on a region that currently doesn’t have any m5 workloads — The purchase recommendation assumes you will move the workload to that other region.
Turbonomic can recommend purchasing RIs for specific workloads or moving to RI tiers that are more appropriate for your applications’ requirements. RI optimization actions are not executed by Turbonomic users. They reflect RI reassignment, which the cloud provider will take care of.
Delete actions relate to removing unused storage such as deleting wasted files to free up storage space or deleting unused storage volumes or devices. The savings that Turbonomic shows are estimates based on the overall cost for that storage. An example set of Delete actions are shown below.
The other actions available are:
- Placement — Place a consumer on a specific provider (place a VM on a Host)
- Configuration — Correct a misconfiguration
- Stop — Suspend an instance to increase efficient use of resources, shown as savings
An ROI before you Buy
Turbonomic ROI is comprised of three components, each relating to the Desired State, and each bearing their own discrete sources of cost savings:
- Operational Expenses
If you are interested in saving money but want to check the analysis before you part with any money, then we offer a proof of value (PoV) phase in which a Turbonomic expert will run an analysis on the 3 items listed above to demonstrate Turbonomic’s ROI in your environment. They can even deliver a specific breakeven date for your investment. The process is simple:
- Scope environment, use cases and measurable outcomes. This is a 2 Hour Session to scope the PoV and agree on outcomes with the exec sponsor
- Business outcomes agreed
- Use cases identified
- Test case/measurement agreed
- We will also hold a 2 Hour Session where the product configuration is discussed:
- Target pre-reqs met by Prospect
- Add required cloud accounts
- Add infrastructure environments
- Add application environments
- Target pre-reqs met by Prospect
- Sign off at both technical and budget holder levels
- Deploy on-prem, in K8S or as SaaS
- Add Targets – discover apps, hybrid cloud & containers
- Walkthrough of Turbonomic running in your environment
- Review Success Criteria
Business Impact Presentation
- A detailed summary of the business and technical gains
- Leveraging data from PoV
In total, we would expect about 8.5 hours total time from you to run the entire process.